The Nuclear Path to 24/7 AI Power: Contracts, Control, and Credit

Opening Insight

AI-driven load is running ahead of grid physics. Short-dated power and generic RECs don’t scale to 24/7, auditable supply . The answer is a power-first operating model built on nuclear-backed capacity —near-term uprates and restarts, with SMR optionality—executed through long-tenor, tolling-like PPAs , disciplined credit and collateral, and a control plane that binds contracts, telemetry, hourly certificates , and settlements.

The downside of waiting is predictable (basis bleed, MTM drift, audit exposure); the upside of acting is compounding (lower cost variance, bankable 24/7 claims, stronger liquidity). We map procurement lanes and timelines (uprates ~2–4 years; restarts ~3–6+ years; SMRs from late-2020s FOAK to 2030s NOAK), codify outage and replacement-power mechanics, and define the ETRM/data modernization required for hourly matching, multi-attribute settlements, and lender-ready controls.

We then translate strategy into Arcelian’s execution blueprint: a control plane, contract stack, governance, KPIs, and a 90‑day plan sequenced to interconnection realities and outage calendars. With that frame, we open Context and Analysis to size the demand–supply gap, timelines, and control requirements that inform the rest of the playbook.

Consequences of Inaction

Delay hardening procurement and risk for firm, auditable capacity, and small cracks become systemic. Timelines and obligations in this market turn slippage into cost.

• Operations/markets: Unfixed queues and congestion widen basis and ancillary/capacity charges; hourly mismatches persist. Transmission upgrades typically take 3–7+ years, and key steam‑path parts can run 36‑month lead times—delays you can’t hedge away.
• Trading/derivatives: Long‑dated load and attribute exposures sit unhedged; FOAK/NOAK schedules slip 12–24 months, and outage/refurbishment risks aren’t modeled, so shape hedges drift and mark‑to‑market discipline weakens.
• Systems/data: Legacy ETRM and risk workflows can’t handle hourly certificates, outage calendars, or multi‑attribute settlements, creating breaks, manual reconciliations, and missed scheduling and settlement windows.
• Credit/collateral: Counterparty and site reviews miss 10–20 year obligations and concentrated single‑site exposure; collateral and margin calls arrive late, and construction‑period LCs aren’t right‑sized to milestones.
• Compliance/audit: 24/7 claims outrun attestation; without hourly provenance and registry integration, expect inquiries, clawbacks on overstated attributes, and restatements when matching falls below thresholds like 98%.
• Financial/P&L: Forced‑outage hours and shape drift force volatile replacement buys; without caps like PPA price + $10/MWh and LD mechanics locked in, balancing and capacity costs bleed through P&L.

Net effect: rising cost of capital, fewer bankable counterparties, and shrinking supplier options just as demand accelerates—eroding siting leverage and delivery windows.

Results of Getting It Right Close

the gaps across procurement, risk, and systems, and power becomes an advantage. With firm nuclear capacity aligned to hourly matching and wired into a modern control plane, trading and operations run faster, safer, and with lower variance.

• Faster decisions and dispatch: Align lane‑specific RFPs, COD milestones, and outage calendars with ETRM modernization and API integration, trimming manual breaks and latency across scheduling and settlements.
• Lower cost volatility: Dependable capacity plus make‑whole/replacement power, forced‑outage caps, and separate capacity/energy pricing reduce balancing exposure and settlement variance via multi‑attribute handling.
• Bankable 24/7 claims: Hourly certificates with registry integration, clear geographic/time rules, and make‑goods for shortfalls create auditable attribution and limit restatements.
• Scalable, right‑sized growth: Lock 500–1,000+ MW from restarts, 20–150+ MW from uprates, and 50–300 MW SMR steps; 200 MW at 92% CF delivers ~1.6 TWh/year to cover two 80‑MW halls.
• Stronger credit and liquidity: Construction‑period LCs, parent guarantees, thresholds/independent amounts, and step‑in rights provide forward visibility and better mark‑to‑market and collateral outcomes.
• More resilient operations: Planned outage calendars, replacement power at defined caps, and reuse of interconnects for uprates/restarts harden availability while a control plane unifies data, workflows, and risk.

Power-First Control Plane

The operating model is power‑first, instrumented by a control plane, and backed by a portfolio that blends uprates and restarts with SMR optionality. Energy and compute become one planning problem, aligning procurement, risk, and technology to deliver firm, auditable, low‑carbon supply.

The payoff: faster decisions, lower cost and higher throughput in scheduling and settlements, more resilience and outage preparedness, better credit/collateral outcomes, and less settlement variance via multi‑attribute contracts and certificates.

• Portfolio mix and timelines: Near‑term increments from uprates in ~2–4 years (about 20–150+ MW/unit), big blocks from restarts in ~3–6+ years (500–1,000+ MW), and modular growth from SMRs—FOAK late 2020s, broader NOAK in the 2030s, 50–300 MW steps. If 300 MW by 2028 is required, prioritize uprates/restarts.
• Control plane and workflow unification: A unified control plane for data, workflows, and risk across front/middle/back office—tying load forecasts, interconnection realities, outage calendars, and scheduling into one decision loop.
• Contract stack and step‑in/outage mechanics: Long‑dated, tolling‑like PPAs specifying capacity and energy, hourly attributes, planned/forced‑outage rights with make‑whole or replacement power, forced‑outage caps, LDs, and lender step‑in with clear default/cure.
• ETRM/hourly certificates and settlements: ETRM modernization to support hourly matching, certificate lifecycle and attestation, outage modeling, and multi‑attribute settlements, with APIs

Event‑driven integration to ISO/RTOs and counterparties.

• Automation and governance: Agentic AI to reconcile telemetry, shape hedges, and trigger actions; rules‑as‑software to encode compliance and bilateral logic; model risk management, surveillance, and attestations for 24/7 claims .

Arcelian Execution Blueprint

Arcelian turns the power‑first strategy into execution by standing up a control plane , a modern contract stack, and the operating model to run them.

We map your load ramp to uprates, restarts, and—when credible—SMRs, sequenced to interconnection reality, outage windows, and bankability.

The outcome is firm capacity with auditable hourly attributes and clear risk ownership.

Architecture

• Control plane unifies telemetry, contracts, schedules, and risk so front/middle/back office share one source of truth and workflow.
• ETRM modernization supports hourly certificates, full certificate lifecycle, multi‑attribute settlements, and outage calendars; integrates via APIs/events with ISO/RTOs, registries, and counterparties.
• Automation and agents reconcile telemetry and trigger actions; rules‑as‑software codify bilateral logic for outages, attributes, and curtailment.
• Risk and analytics include ML demand/outage forecasting, schedule‑risk and fuel‑supply scenarios, construction/credit event modeling, and mark‑to‑market for multi‑attribute contracts.

Roadmap and sequence

• Convene a 90‑minute working session with CIO, COO, CFO, CRO, and Head of Trading; deliver a 90‑day plan to stand up the control plane and contracting strategy.
• Run lane‑specific RFPs now; map load to uprates (≈2–4 years; 20–150+ MW per unit), restarts (≈3–6+ years; 500–1,000+ MW), and SMRs ( FOAK late 2020s; NOAK 2030s), adding 12–24 months optimism risk to FOAK schedules.
• Fund the control plane and contract stack in parallel with siting; place transmission/interconnection (3–7+ years) and a 36‑month steam‑path lead time on the critical path.
• Sequence tranches and term sheets (10–25 year tenor) with outage calendars, step‑in/LDs, and hourly attribute delivery to protect COD windows .

Governance, rules, and data model

• Establish model risk management , attestations for 24/7 claims, and surveillance to reconcile sustainability statements with underlying telemetry.
• Codify bilateral rules as software: geographic/time rules for hourly certificates, make‑goods and clawbacks, forced‑outage caps, replacement power and basis allocation, and change‑in‑law handling.
• Core data objects: hourly certificates, outage calendars, capacity/energy/attribute settlements, credit/collateral terms—each with audit trails and registry/API links.

KPIs that matter

• Hourly matching ≥ 98% monthly with make‑goods within 60 days; validated, auditable certificate trails.
• Forced‑outage hours ≤ 2.5% annual cap with planned work in shoulder months; on‑time milestone delivery vs COD windows.
• Reduced settlement variance via multi‑attribute handling and node/busbar

Delivery Mechanics and Collateral Design

Lower collateral drag through forward visibility and covenant design.

• Coverage versus load: track MWh against targets such as 200 MW at ~92% CF ≈ 1.6 TWh/year .

Human and Organizational Actions

• Stand up a cross‑functional nucleus across trading, scheduling, credit, treasury, sustainability, and architecture; refresh RACI so procurement strategy, hedge design, and attestations are owned and sequenced.
• Role clarity: CIO (systems/data and ETRM modernization); COO (operations and outage execution); CFO (credit/collateral/liquidity); CRO and Head of Trading (risk limits and hedge design).
• Align incentives to lifecycle value:
  • Negotiated outage terms
  • Reduced settlement variance
  • Lower collateral drag
  • Validated hourly claims
  Build hourly accounting and outage preparedness into day‑to‑day operations.

Trade‑offs to Manage

• Speed/scale/bankability/interconnection: uprates (fast, smaller blocks), restarts (bigger blocks), SMRs (modular, later); factor FOAK optimism risk.
• Wires and site: plan 3–7+ year transmission upgrades; reuse ties for restarts/uprates where feasible.
• Credit/collateral and change‑in‑law exposure: size security, margining, and step‑in/LDs to 10–25 year obligations.
• Basis and attributes: lock node/busbar delivery, hourly certificate rules, curtailment and replacement‑power mechanics to protect hourly claims and P&L.

Secure Firm, Auditable Capacity

AI‑driven load is outrunning grid reality, and short‑dated buys plus generic RECs can’t square uptime, price stability, and hourly clean claims. The practical path is to choose the lane that can actually firm your timeline—uprates in ~2–4 years for 20–150+ MW increments, restarts in ~3–6+ years for 500–1,000+ MW, and SMRs scaling from FOAK in the late 2020s to broader NOAK in the 2030s—and contract with tight credit, outage, interconnection, and attribute mechanics. Get it wrong and you strand capital, widen basis risk, and invite scrutiny; get it right and you lock in low‑variance costs, credible hourly attestation, and siting leverage.

Strategic takeaway: adopt a power‑first operating model, run lane‑specific RFPs, and fund the control plane and contract stack in parallel with siting to secure firm, auditable capacity at scale.

Launch Your 90‑Day Plan: Call to Action

Arcelian turns uprates, restarts, and advanced‑reactor options into bankable supply for AI‑scale load—delivering firm, auditable capacity and credible hourly matching on your timeline.

• Contract and portfolio design for uprates, restarts, and SMRs with dependable capacity, outage calendars, replacement power, and attribute make‑goods tied to hourly certificates.
• Credit and collateral engineering to size 10–20 year exposures, set covenants and triggers, and align liquidity with construction and outage risk.
• ETRM and data modernization enabling hourly

matching, certificate lifecycle, multi‑attribute settlements, and event‑driven integration with ISO/RTOs.

Operating‑model and governance alignment across front/middle/back office with ML‑driven forecasting and schedule‑risk scenarios that withstand audit.

Convene a 90‑minute working session with your CIO, COO, CFO, CRO, and head of trading so we can map target load growth, siting options, and candidate uprates/restarts/advanced reactors, then deliver a 90‑day plan to stand up the control plane and contracting strategy for firm, auditable capacity at scale.

ETRM & Platform Modernization: Choosing the Right Path for 24/7 CFE

Most firms face three viable modernization strategies: (1) replace the ETRM core, (2) augment it with a power-first control plane, or (3) federate via a data/event backbone while selectively refactoring high-friction modules.

The right choice turns on time-to-compliance for hourly matching, ability to support multi-attribute settlements (carbon-free, locational, vintage), native outage calendar handling for nuclear PPAs (uprates, restarts, SMRs), and the integration roadmap to ISO/RTOs and registries.

Prioritize an ETRM architecture that exposes clean APIs, extends the product and contract models without customization debt, and guarantees lineage from meter/telemetry through settlement and credit.

This section reinforces the post’s thesis: a power-first control plane plus targeted ETRM modernization is the prerequisite to firm, auditable 24/7 CFE at AI scale.

Augmentation typically delivers the fastest path: deploy an event-driven control plane that normalizes telemetry, schedules, and certificates; orchestrates hourly matching; and posts enriched events back into the ETRM for pricing, risk, and settlement.

Trade-offs include dual governance and tighter SLOs across systems, but you avoid multi-year rip-and-replace risk.

Full replacement can simplify the estate and unify controls, yet requires proven adapters to ISO/RTOs and registries and a data model that natively encodes hourly attributes—rare in off-the-shelf platforms today.

In all paths, agentic AI belongs on top of governed data and controls: monitor outage calendars, pre-match hourly certificates, flag credit exposure spikes, and route exceptions—within policy guardrails and with immutable audit trails across front, middle, and back office.

Recommended sequencing and outcomes:

• Phase 1: Stand up the event hub and canonical data model; connect ISO/RTO, registry, and telemetry feeds. KPI: >95% automated data ingestion with lineage.
• Phase 2: Implement hourly matching and multi-attribute settlement services; integrate outage calendars. KPI: <1% unmatched hours; T+1 provisional settlement.
• Phase 3: Expose APIs to ETRM risk, credit, and invoicing; deploy AI-assisted exception handling. KPI: 50% reduction in settlement cycle time; zero material audit findings.

Frequently Asked

Which contract terms belong in long‑tenor nuclear PPAs to stabilize costs and support hourly 24/7 claims?

Specify capacity and energy separately; require hourly certificates with registry integration and clear geographic/time rules; include planned/forced‑outage rights with calendars, make‑wholes or replacement power at defined caps (e.g., PPA price + $10/MWh), forced‑outage caps, LDs, and lender step‑in with clear default/cure. Lock node/busbar delivery, basis and attribute allocation, and multi‑attribute settlement mechanics so balancing costs and attribution disputes don’t leak into P&L.

How do uprates, restarts, and SMRs compare on timing and scale, and how should we sequence procurement?

Uprates: ~2–4 years, about 20–150+ MW per unit; Restarts: ~3–6+ years, 500–1,000+ MW; SMRs: FOAK late 2020s, broader NOAK in the 2030s, 50–300 MW steps. Add 12–24 months optimism risk to FOAK schedules. If you need ~300 MW by 2028, prioritize uprates/restarts. Plan for 3–7+ years on transmission/interconnection and up to 36‑month steam‑path lead times. For scale, 200 MW at ~92% CF is ~1.6 TWh/year—big enough to magnify timing and balancing risk if supply slips.

What is a power‑first control plane and why is it the fastest path to auditable hourly matching?

It’s a unified operating layer that normalizes telemetry, schedules, and certificates; runs hourly matching and multi‑attribute settlements; and integrates outage calendars—posting enriched events back to your ETRM and into ISO/RTOs and registries via APIs. Augmenting the ETRM with this control plane typically delivers the quickest compliance: target >95% automated data ingestion with lineage, <1% unmatched hours with T+1 provisional settlement, and 50% faster settlement cycles with validated audit trails and ≥98% monthly hourly matching (with timely make‑goods).

Trend Watch

The market is coalescing around a simple mandate: pair a power‑first control plane with nuclear‑backed procurement to lock firm capacity for AI and verifiable 24/7 carbon-free energy. That means structuring tolling‑like nuclear PPAs with separate capacity/energy, make‑whole and replacement power, and hourly energy certificates wired to registry integration. On the platform side, ETRM modernization must treat hourly matching, outage calendars, and multi‑attribute settlements as native—not bolt‑ons—so trading, risk, and sustainability can operate from the same ledger. Why it matters now: nuclear uprates and restarts can land meaningful MWs on near‑term horizons, while small modular reactors (SMRs) shape modular growth as FOAK turns to NOAK in the 2030s. The winners will pre‑wire ISO/RTO integration, certificate lifecycle, and credit telemetry into the control plane, then let AI in

ETRM automate reconciliation, forecast outage exposure, and trigger hedges within policy. That reduces basis bleed, MTM noise, and audit risk as data center demand accelerates.

Choosing the right modernization path comes down to three proofs:

• Data and interoperability: Can your event backbone reconcile meter data to certificates hourly with immutable lineage and pass enriched events back to risk analytics?
• Contract intelligence: Can systems price and settle tolling‑like PPAs, enforce outage windows, and apply change‑in‑law logic without spreadsheets?
• Credit and liquidity: Can you size 10–25 year collateral for nuclear PPA tranches and surface exceptions in T+1?

Firms that answer yes will bank credible 24/7 claims, unlock siting leverage, and compress settlement variance as AI‑scale load ramps.

Closing Insight

AI load has shifted power procurement from opportunistic hedging to infrastructure‑grade risk management, where resilience is coded into contracts, systems, and credit from day one. The edge now goes to operators who pair nuclear‑backed capacity (uprates, restarts, SMRs) with a power‑first control plane and ETRM modernization that treat hourly certificates, outage calendars, and multi‑attribute settlements as native flows—not bolt‑ons.

That stack compresses basis bleed and mark‑to‑market noise, hardens 24/7 attestations, and converts volatility into governed automation as AI monitors outages, pre‑matches certificates, and routes hedges within policy. The strategic move: run lane‑specific RFPs, lock tolling‑like PPAs with make‑whole and replacement power, and fund the control plane and collateral design in parallel with siting—so capacity, compliance, and liquidity scale together and the next MW added strengthens cost stability, audit posture, and siting leverage.

Partner with Arcelian

Partner with Arcelian to convert AI‑driven load into firm, auditable capacity—sequencing uprates, restarts, and SMR optionality with tolling‑like PPAs, hourly certificates, and outage mechanics your auditors and traders can trust. Our team stands up a power‑first control plane, modernizes ETRM for multi‑attribute settlements, and engineers credit/collateral and step‑in rights so 10–25 year obligations are bankable and MTM discipline holds.

Connect with our team to explore a 90‑day blueprint that maps target MW, interconnection realities, and liquidity constraints into an execution plan that stabilizes P&L, hardens 24/7 claims, and strengthens siting leverage.